JOURNALOFTHECANADIANSOClErY OF EXPLORPlTlON GEOP”“S,C,S,S
VOL.19, NO,1 ptc. 1983,.P. 16-33
USING A VERTICAL SEISMIC PROFILE TO INVESTIGATE
D. HAMPSON’ AND L. MEWHORT~
ABSTRACT
A major benefil of the Vertical Seismic Profiling Method
(VSP) sternsfrom its ability 10distinguish multiple eventsand
identify the generatingsystemunambiguously.In this study, il
VSP was recordedin an areaof we~rernCanadawhere long-
period multiple are a particular problem. One of the ohjec-
tives of the study was to clarify the multiple problem at the
zone of interesf and evaluatemethodsol multiple attmuation
on the seismicdata. Effective attenuationmethodsin lhis area
would gently assistsrismic interpreters.
By separatingthe VSP intu its up-goingand down-goingwave
cumpunznls,the majorprablemwu identified 8* a first-order
multiple from a shallow event. followed by il weakerprg-leg.
In order to makemeaningfulcomparisonswith seismicdata.
thedown-goingwaveletcharactermeasured hy thewell eeophone
was analyzrd. This analysis shr,wedmajor variations in hnlh
phaseand sprctral bandwidth. However, a seriesof control
waveletsmeasuredby a geophonea, the surfxe indicatedthat
these variations wrre due mainly lo changing shot-hole
conditions. Hence it was assumedthat the ~avrlet character
did not vary significantly with depLh and ~hr VSP was
deconvolvedby using the individual down-cuing wavrlets.
The excellent tic betweenVSP and surface seismicd&a, as
well asrem-phasesynthetic data.cunfirmrd the validityofthc
deconvalution procedure.
To minimirc multiple interference on the seismicdata. wo
techniqueswereemployed.One methodusedthe VSPitselfto
designthe attenuation <,peratur,while the other methudused
differential m~vecutin thr F-K plane. In eachcase.the result-
ing seismic data were evaluated by comparing the seismic
responsewith the VSP response.
INTRODUCTION
In March 1981, a Vertical Seismic Protile was shot by
Hudson’s Bay Oil & Gas Ltd. in an area of western
Canada where long-period multiples are a particular
problem. It was hoped that by studying the multiple
wave train and primary response separately, the multi-
ple generating system could be understood and various
multiple removal schemes evaluated.
This paper describes the acquisition, processing and
interpretation of the VSP and also evaluates two multi-
ple removal schemes on a template seismic line.
’ Veritas Seismic Ltd., 615 3rd Ave. S.W., Calgary, Alberta
’ Formerly Hudson‘s Bay Oil &Gas CompanyLimited. Now CanterraEnergy Ltd.. 555 4th Ave. S.W.. Calgary, Alberta
16
A MULTIPLE PROBLEM IN WESTERN CANADA
VSP FIELDAC~UISITION TECHNIQUE
Energy was provided by shooting 0.25 kg dynamite
charges in a 15-m deep, 0.8-m diameter, cased and
cemented hole. A 12-m water tamp was maintained
throughout the survey. Energy was received by a wall-
lock well geophone which was locked in position for
every shot at 15-m intervals from total depth to surface
casing. The well geophone had a natural frequency of
IO Hz and a frequency amplitude response within 3 db
from 8 lo 300 Hz. The cable was slacked 8 m at each
shot level so cable waves would not be transmitted.
The shot-hole was located 65 m from the well. It was
felt that as this area did not have a strong ground roll
problem, tube waves would not be significant and thus a
short shot-hole to well distance could be tolerated. This
proved to be the case as only low-amplitude tube haves
were observed on the VSP (Fig. I).
As the data quality was good, and to minimize rig
standby time, each level was shot once. Twelve hours
were required to shoot 130 levels. To preserve a quiet
environment, all motors on the well site were shut down
for the duration of the survey.
The signal was recorded on a four-channel multigain
ranging digital recorder at a I-ms sample rate. One
channel was used to record the measurements from a
sulface reference geophone. Several surface reference
geophones with varying natural frequencies and damp-
ing factors were used at different times during the SW
vey to test their responses. This was unfortunate, as it
complicated the use of the reference signal in determin-
ing changes in the source wavelet over time. The prob-
lem was alleviated by dephasingall the reference geophone
signals at the processing stage.
The 15-m vertical spacing was selected to ensure that
a relatively broad-band response would be preserved
after F-K filtering.
TEMPLATE SEISMIC
Thetemplateseismicline, tobeshowninlaterfigures,
was shot after the well was drilled. These data were
 “ERTlCAL SElSMlC PROFlLE DISTINGUISHES MULTIPLES 17

1965 DEPTH (METERS) 240 symmetrical split spread, recordingfiltersofout-I28 Hz,

- and a l-kg dynamite charge in a 14-m shot-hole.
I
The processing sequence included exponential gain
recovery, spiking deconvolution, normal moveout
correction, automatic statics, bandpass filter and stack.

PROCESSING

One of the major benefits of the VSP method stems

from the fact that, as the receivers are measuring sig-
nals at depth, energy is recorded while moving down-
ward as wellas upward. Asaresult, the events recorded
on the VSP form a simple pattern as shown in Figure I
- all events dipping from right to left (such as the first
breaks) came from energy moving down through the
earth, whereas the events dipping the opposite way
were moving upward when recorded. The interpreta-
tion of the various events is straightforward and has
been discussed by numerous authors (e.g., Kennett,
Ireson and Con”, 1980). Basically, if the offset from
horehole to shot is small and the reflecting horizons
have small dips, all the downward-dipping events are
strictly parallel and have a shape that follows the aver-
age velocity as a function of depth. Similarly, the
updipping events are parallel and have a dip with the
Fig. 1. Vertical seismicprofile beforeprocessing. same magnitude but opposite sign from the down-dipping
events.
Figure 2 shows the interpretation of the major events
recordedwitha96-channel instantaneous floating-point on the VSP. All the down-dipping events are multiples
recording zsystem with a 2-ms sample rate. Recording except for the first breaks. The up-dipping events can
parameters were: a group interval of 33.5 m, a shot he either primaries or multiples-the rule being that if
interval of 134 m, offsets from 33.5 to 1609 m on a the event extends all the way to the first breaks it is a

1 2 3 A 5

1.
OlREc
2.ARRIV
pm/N
PRlMA
REFL
(“P-
‘\i

3.
FRST-
MUL
(Ww
V 4. FIRSTCIRDER MULTIPLE (UP-WAVE)
5. PRIMARY REFLECTION (“P-WAVE,

Fig. 2. Interpretationof VSP.

 D. HAMPSON and L. MEWHORl
1. INPUTVSP
Fig. 3. Producingup-wave section
primary (as in event 2); otherwise it is a multiple (as in
event 4). The deepest primary involved in generating a
givenup-goingmultipleisidentitied hytindingthe termi-
nation point of the multiple event and tracing vertically
upward through the VSP to the first breaks, which
should coincide with the termination point of a primary.
In Figure 2, event 4 is a multiple generated between
event 2 and the near surface.
For analysis, it is convenient to separate the VSP into
two parts, one of which contains only upward-moving
energy, and one of which contains only downward-
moving energy. This is achieved by F-K filtering as
shown in Figures 3 and 4. In Figure 3, the iirst breaks of
the VSP are aligned by applying a series of static shifts.
In the process, the other down-going events are aligned
as well, since they are all parallel. In Figure 4, an F-K
filterhasbeenapplied. whichsuppresseszero-dipenergy.
and the original static shifts are removed twice, with the
result that the up-going events are aligned. Similarly,
the up-going events may be suppressed, leaving the
down-going events aligned.
2. STATICSHIFTS
TO ALIGN FIRST
ARRIVALS
The two resulting displays are shown in Figures 5 and
6. The horizontal axis of the up-wave section (Fig. 5)
measures the depth of the receiver, and each trace
shows the seismic response that would result if seismic
data were recorded with the receivers at that level (and
only upward-moving energy recorded). The multiple at
1000 msisclearlyvisihle,as wellastheprimaryeventat
1050 ms. It is also clear that the reflection quality ofthe
primary is affected by the presence of the multiple.
Figure 6 shows the nature of the multiple problem.
The interpretation of the down-going wave section is
that each trace shows the total incident wavelet at every
level. By studying the wavelet change from trace to
trace, we can track the change in wavelet character
with depth (assuming a constant shot condition). The
first breaks in Figure 6 can be thought of as the basic
seismic wavelet, while the later events are the multiples
that will form the tail of every primary reflection below
the receiver. In this case, most of the multiple energy is
concentratedabout msafterthefirstbreaks,indicat-
ing that most of the multiples on the up-wave section (or
 “ERTlCAL SEISMIC PROFILE DISTINGUISHES MULTIPLES
3. F-K FILTER
TO ATTENUATE
ZERO-DIPENERGY
Flg. 4. Producingup-wave section.
seismic data) are caused by the single extra bounce
between the primary at 500 ms (top of Wahamun) and
the near surface. Other smaller-amplitude events on
Figure 6 between 100 ms and 600 ms correspond to
shorter-period peg-legs possibly set up between the
Wabamun and later events. A multiple that is set up in
this way could not be measured as a down-going wave
by receivers above the Wabamun, so that such an event
should terminate at the Wabamun depth (534 M) on the
down-wave section. Although some of the events on
Figure 6 (e.g., around 460 ms) do appear to terminate in
this way, it is not clear at exactly which depth.
In order to study the basic wavelet character, Figures
7 and 8 show the first breaks (without spherical diver-
gence compensation). Also shown is a plot of the ampli-
tude spectrum of each wavelet in the range O-100 Hz.
The analysis of this display is an important part of the
processing, because it is essential to bring the up-going
wave VSP as close as possible to zero-phase. On exam-
ining the wavelets, it is apparent that they fall into three
distinct groups: 1) the wavelets measured above the
19
4. STATICSHIFTS
TO ALIGN
UP-GOINGWAVES
Wabamun, 2) the wavelets measured below the Waba-
mun, but above the level marked “characterchange” in
Figure 8, and 3) the wavelets measured below this level.
It is important to determine whether the changes observed
here represent real changes in the wavelet character as
it moves through the earth, or whether it is the shot
condition that is varying.
First of all, the wavelets measured at the deepest
levels are apparently of broader bandwidth than the
wavelets measured above - that is, opposite to the
expected wavelet changes due to absorption orinterbed
multiple generation. Apparently, the sudden change
observed at the deeper level is due to a change in the
shot condition itself. Figure 9 tends to support this
interpretation. A geophone located on the surface close
to the bore-hole was used to measure the direct arrival
from some of the VSP shots. The surface geophone was
either a 30.Hz or a 50-Hz phone as indicated in the
figure, whereas the down-hole geophone was a IO-Hz
phone. In Figure 9, the two sets of wavelets are shown,
after the appropriate phase compensation for the differ-
 D. HAMPSON and L. MEWHORT

1965 DEKH (METERS) 240

c-
,,,, :‘i”N:‘y 8,
:: ;,;;,,;l,l:!~:.,!,,~J:.,.::,:..,:~
,,’ 18, ,,: ;, i,:;,!,,: ,! ,,
,,..!~>:iLi.;!::,~;;..‘:.li
,,.1,,!11:1;:!;l:l:l:ll:?;?i.
~~&:~.,,~dv 4-
::::;::;;~:~::~“:‘;~;~,~;;,~,,.,’,,,,
I,:/,,/,,;,(,,),,,a,‘,)l,,,ll.il:,,,,,ri(m~sjfi,~,lil~
,,,,,Ijl,::l,,.:.!ij,!!-~.
,:.‘:.l:.i,:i::,):::,::~i s’!,;,,m,!(,,,i(8:(,,(;I,:;,,I,,,,,,,m,,:
#,,! ,I,,,((.!,,!(I,,I,‘,:,~(,,:,,,,/~,,,,‘I!,,,(/,
:~~:_:i’;~li:i~;i~i~:~~i~~_~::~:~
~~;~~:~::~‘~~~~:I~;~:!~J~~~~:~I,~~~~;~ ~r~:i~I:~ ~zr;’ ?~~:~‘;~‘f’~;‘~~~
, ,,,, ;, ,‘!
il::!‘::II’i! ,,‘!z,:!‘,8’8”’‘: I,:#:,‘,,!‘,#,I’‘i’,‘:, !,ls,,mI’,::‘,:
,~,,“:!ll!!il;~‘:‘,.,,‘,,‘, /~11.1~1~.1,
///I/I,,,l/,s/i~,;
,,,/,:i:,;:;,.:i,,,,,:/,I, ,,,,
~,f’:f~s
~~I~,W/////,,// //1/,//,,,/,/,,,,,,,,,,,,,,,,,,,,,,,,~,,,,
,,ll.l,,lllti,i;:;i:::::;!::li!,il:::;l;;,:::;~”

Fig. 5. Up-going wave section. Fig. 6. Down-going Wave sectIOn

FREQ”ENCY(HZ)
0 50 100

Fig. 7. Down-going waves: depth 240 m lo 1095 m. Fig. 0. Down-going waves: depth 1100 m lo 1965 m.
 VERTICAL SEISMK PROFILE DISTINGUISHES MULTlPLES 21

1200 M

MEASURED WITH
30 HZ PHONE: i
I
CHARACTER _
CHANGE

MEASURED WITI-
50 HZ PHONE$
I; ~j::
,L

,&.
A,&.
I*
--
LLee.-,

__._

,Ah-*.-,
J Y
~_
__._~.
I------.
1.::::
--
-.
--
-~~,
~.~~,
~-

I
A .A -- .-
j;.;: I;;~
-1 L. -_ ~~.
,.A _. .-
--1* _.._ :
1965 M
0 500
b
TIME(ms)

Fig. 9. Comparison between surface-measured waves and VSP W~YBS

ent geophones. Also shown is the cross correlation
between each pair of wavelets and, finally, the same
cross correlation band-limited to the range 45 to 90 Hz.
It is clear that both sets of wavelets exhibit the same
change in character at the same location. The cross
correlation, on the other hand, shows that, although
there is a measurable phase difference between the two
wavelets in each pair, that difference tends to remain
constant as the two wavelets change. The band-limited
cross correlation shows this consistency even more
clearly, as the changeinamplitude spectrumisnolonger
visible. Asit happens, theVSPwasshotwiththereceiver
moving from the bottom of the hole upward, and the
effect we are seeing at depth is a change in the shot-hole
condition with time. This interpretation is neatly coo-
firmed by the fact that several widely spaced calibration
shots were tired while the geophone was being lowered,
and these wavelets show the same character as the
deepest wavelets. One such wavelet is the ninth below
the level marked “top of Wabamun.”
The explanation for this effect is not entirely clear,
but apparently some sort of inelastic deformation of the
region about the shot-hole has occurred after about the
first thirty or so shots. According to Ziolkowski and
Lerwill (1979), the bandwidth of the basic pulse varies
inversely with the size of the sphere of anelastic defor-
mation about the shot. The changes observed here would
be consistent with the hypothesis that this region of
anelastic deformation had suddenly expanded and stabi-
lized at a new value.
The wavelets measured above the Wabamun are appar-
entlyofadifferentcharacterfrom thosemeasuredbelow.
However, the magnitude of the change is so extreme as
to make us question the validity of these shallow wavelets.
As can be seen in Figure I, a great deal of high-frequency
noise exists in this region, while a close examination of
22 D. “AMPSON and L. MEWHORT

FREQUENCY RESPONSE
TIME RESPONSE

AVERAGE
WAVELET FROM I II
ABOVE
WABAMUN

4I 81 :?I

FREQUENCY (HZ)

AVERAGE
WAVELET FROM Bes / J\ ,/‘?-:A-.

L
BELOW P 858 m
WABAMUN TIME(ms)

1b 81 ISS

FREQUENCY (HZ)

Fig. IO. Average wavelets calculated from first arrivals

FREQUENCY RESPONSE

TIME RESPONSE
>II

AVERAGE
a0 I $8
WAVELET IJI aa
FROM ABOVE
WABAMUN TIME (ms)
~,,O~~ n-” . 86L
II 8. ,a*
FREQUENCY (HZ)

I 88

AVERAGE
WAVELET I .I 8$6
FROM BELOW 0 ZII~
WABAMUN w TlME(ms)

.I II r\-* I II L
?I 80 12.
FREQUENCY (HZ)

Flg. 11. Average wavelets calculated from first-order multiple.

 VERTICAL SElSMIC PROH‘E DlSTlNGLr,SHES M”LT,PLES
the first breaks on the inside traces revealed that digital
clipping had occurred in many places. Any change in
wavelet character on passing through the Wabamun
should be observed not only on the first breaks, but also
on the first-order multiple on the down-going VSP.
Figures IO and 11 show the average wavelet measured
above the Wabamun compared with the average wave-
let measured below. The average wavelet from below
was calculated by summing the wavelets from the level
marked “top of Wabamun” to that marked “character
change.” Figure 10 shows the result of using the first
breaks, while Figure 11 shows that of using the first-
order multiple. It is clear from Figure 1I that, aside
from a small high-frequency attenuation, the wavelets
from above and below the Wabamun are practically
identical. For this reason, it was decided to discard the
wavelets measured above the Wabamun as unreliable.
Finally, the large group of wavelets from the middle
traces of the VSP show a stable wavelet phase with a
gradual attenuation of the high frequencies. The very
slight degree of high-frequency absorptionevident below
the Wahamun is consistent with the findings of other
investigators such as Huge (1981), who reported atten-
uations of less than 0.1.db wavelength in rocks that
were predominantly limestone and shale.
Because the wavelets showedlittlechange with travel-
time, it was decided to assume that the down-going
wavelet at each receiver level was representative of
that entire trace and to deconvolve each trace with its
own wavelet. As the wavelets measured above the
Wahamun were considered unreliable, the average wave-
let measured below the Wabamun was used for these
traces. Figure 12 shows the resulting up-wave section
Fig. 12. Final dephased up-wave VSP.
23
1965 DEPTH (METERS) 240
-
I I
“~~‘:” 8:‘:~:ii~‘~~-~;~~,:‘,~‘:, :y;: ,, ,, ,, ,,,,, ,,,‘:,‘,, ,,,,, i, ,, :: ,~,, ,, ;; ,,m,,: I,,
,,,, ,, mm,,;; m,;< f,
!~,~/,,,,,,,,,!!!!,!
::,::,,,,!/,(/,,,I,/,!,,/,,I,#~,~
,I/,/mm, /WI/,lS,,
I/,,/,I,,illllill,li,t,i,(:,,ii,II:’ :‘,N’,:::,,“88
Yc
/,,/ ~,,,//,,;//,,)/;,;
,,,,.,,,,.,,.,.,/,.,,,,_,,,/,,,,,,, II,!,1,, r ,1,,111,,,,,:)l,.*lltw,m~,ww.‘..‘.:!:.,.“.“....,,...,,!~,~!!!,i,~
,,,,,:,,, 8,,,‘,I ,,,,Il!f:1,:!~~s#:!:::I
8;:;y,,i;,:-,;;, ;;,;; ,f ,: ,;, 8:;,,:; 1,,,; ,: :; ,, ,;, ,, ,N1;,, ;: it’,, ‘,,,,,//‘I”‘,,,/~,~,
b,,,,,,,, ,,,,,,,,,,,I,,I,/l,,l,i,~~:,~i~~,:,.,,I,,I~,,I,,II,,/,,,!,,,~,/,
:..:~::::.:;,.:..:::~:;~.~~.,~.., ,.,....,,,.,
::,:/,:I,,~.,,::,:,,~,,..~,I.~..~,.~.......
1; :,,:i,,: ,;,;y ,,,,:i
~~~ iiii’ciiiiii;liliiri~~~~~~~~~~,riii::illi:ll;;:~iilllli:l~i:~lii;i::iI
!!,,:,,,ti~ l.,ll,ii~clwil.li,l,i,,,i,~lm”~,,~:,:,;~~
,,,,:;;,;:;;:;;;:;;:,;,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,..,,..,,,,,,,,~,,,,111.1.111 i//iii I....Il::i% “’
!~~.v!~iiii~i”ii,~~,~!,~,~,,,
-4
~,l,llillr!l!llii,snrli’l’i!,,I, I 1~llllf~lN
,,.,,,,,._,,n!,,,“dl,l,,,),i,,,,,!
‘I :...,,,>“, y~,“,u,>,m>w
,,, ~~,“,,‘:,~‘;‘“‘,‘,,“‘,“,’“““““““11”“,,,,~“.~1.1I~~.!~~il~II~Il,l,lliilyllllil:ll~“~~~~,.~~l.,l
,,,,,, ;/!,,,,,:,!,,,,,,,;;;/!,!:~,I,.,i/!!:/,,/!~!I::/~I!:~,iill,,jl,~l~’!~:,’,
;,:I;:,;~;iil!!‘:;,;~.“::.~,,,‘,, ,,,i,,,,,,,,,;.,‘,,‘,,,,,,“‘,,,ll,llllll!!l!,,11//1,,,,,,,,,,lll,lll,.ll,ll,!,,,!
,,~~~ 2
~~:i:i,,:,::i~~.;:,:j~,;::il;i~~ ;~~:~~::~~,~:,I’~,~::,,‘~~o~~/,,:,~:~~:i,,/,~/,~~~~I~~~~~~I~~,”/~~~~,/,~~/~ ,;;,
ii::~,,I.:!l.:l:~Il:i,;~l~,I~ ,;/,/,,,,/;,,!,/I,,,;:::,,!,i!!!t//:N//, ,,,11,,,,,,11, I:,Illllliii. .dm!lJ.,
,~~
,,,” ,,,,,,,,,,,>..,,..,..>. ,,,v ,,lllllll,l))llll,l!i ,,,,,,,,,,\,:,.si/,,,,,,,j,,,,,,~,1,l1,)*1,11,11)11111111))
!;,l’liiliili,;i ,,/,,~//,,,,w,, I I’““‘“’ ,,1,!,!! ,~ ~.~ ,,,,,.~~.~~~~~~~~ ,., I
111!!~!~1~~,,~~~,I’~~i’~~1~/1~~I~I~/~~,///~,~~,,~,li~‘~~~.~’.,1
“’ ;j:::;:::::::::;;:;,:iJ,;,,::::~?*:,I,~,,,l~.,,,l...,,,,,,,,,l,,l #‘I,li,,,, *,.,,.,,>1,%,1,,# ,,,,,,*tl*.wl-
~l/~,,.ll,,~~l,‘,,~I,,,, ,,
I !) , ,/,t jt/;llll~,l,,lll,~,,I /,,,,,,,/,,.,“,,.,,,,,,,,,,,)11))))))1111110)11,,,,,,
,.~7”“.“‘““““,,,,,,,,,,,~
“,,,I,,.,,,,,,,....,11,11 -, ,,I,f”,! ,,,,,,,!,,!,,,WN~,,~ ,111,
!//!/,,/,1,!1~!1,,1,ll~i1iillill!,!1;:~
::,,:, ...,.,,t,,,,, .,, ,,‘;:!““lrl:!,l”““‘““‘:““‘:“““““,
-.-...*Y.,,.l*~“~~~)))II l,,///ll/l,,,lll,,l!~l)I:I:)(:
,,I,,,” ,/,,,,, ,,,
::!!::l:lll’!::::::::::::::::::”/‘,,;l//:l1,111, IIIIw!i:;Ji/,,,,~l/,jil!!““,,‘,,,,,,,/,,/,,:,,,,‘,:‘!F :‘I
Fig. 13. Down-going waves after dephasing.
after using each down-going wave to correct the trace to
zero-phase while leaving the amplitude spectrum un-
changed. The Wabamun reflector now appears as a
single positive reflection coefficient, as it should. The
dephased down-wave section is shown in Figure 13. It
is interesting to note that while the first breaks are
zero-phase, thetirst-ordermultipleisnotexactlyreversed
in polarity. This deviation from reverse polarity mea-
sures the phase change in the wavelet due to the two-
way travel path between the Wabamun and the near
surface.
VSP INTERPRETATION
A major problem for interpreters in the area under
investigation has always been the presence of strong
multiples at reflection times ofinterest. Figure 14 shows
the lithology in the vicinity of the well, and it is clear
that very strong reflections will be generated from the
shale-carbonate boundaries. In particular, the upper
boundary (top of the Wabamun Formation) has an espe-
cially high acoustic contrast with a reflection co&i-
cient of about 0.4.
In order to compare the VSP with surface seismic
data, composites of the VSP traces over specific zones
were formed. Figure 15 shows one such composite. The
narrow area following and paralleling the first arrivals
has been called the primary zone because no multiples
generated between the Wabamun and the near surface
are present in this zone. As this mechanism is responsi-
ble for most of the multiple energy, stacking the traces
within this zone produces atrace that is relatively multiple-
free. Similarly, as shown in Figure 16, stacking those
24 u. HAMPSON and I,. MEWHOKT

traces recorded above the Wabamun gives a good approxi-

mation to zero offset surface recorded data. since all
primaries and up-going multiples generated by the
Wabamun or lower primaries are included in this stack.
By correlating the two composites with the seismic
data, it should be possible to distinguish multiples and
primaries unambiguously.
On the left side of the up-wave VSP (Fig. 161, two
deep primaries can be observed: one at I. I5 s and the
other at 1.I8 s. The amplitude and structure of these
events are critical clues to hydrocarbon accumulations.
Both events deteriorate on the right side of the VSP as
more multiples are included. Apparently these interfer-
ing multiples are generated by primaries below the
Wabamun. The primary at 0.65 s, for example, has
createdamultipleat about I. IS sasitcrosses theappro-
priatc lcvcl. This result is not obvious to interpreters
(from examining conventional common-midpoint or
common-offset seismic stacks). In the next section, this
interpretation is corlfirmed by showing an improvement
in the primary response at this level after multiple
attenuation.
One of the objectives of the survey was to eVahK3tc
the reasons why synthetic seismograms created from
%Z sonic logs in this area did not correlate well with the
rzd SH*LIE lower part of the seismic section. Figure I7 shows the
/ANHYDR’TE VSP primary trace correlated with the primaries-only
WLOMlTE synthetic seismogram after artilicial checkshot adjust-
L,MEST*Nt ments to match event spacings and identical filtering.
r<-. IGNEOUS The excellent tie illustrates not only the correctness of
1‘. :.: ROGK
our approach in generating a primaries-only VSP
composite, but also the good performance of the sonic
log. Apparently, much of the difficulty in matching
Fig. 10. Stratigraphy of VSP well synthetics with conventionally processed seismic data

PRIM*w
DEz=RS1 ZclNECDMPOSITE

Fig. 15. ~pwave VSP. dephased and filtered 318 - Q5!125 Hz. Fig. 16. up-wave VSP, dephased and filtered 318 - 951125 Hz
 VERTICAL SEISMIC PROFILE DISTINGUISHES MULTIPLES 25

PRIMARY
ZONE
COMPOGITE
v li”‘,,‘~~:::~~:~,-:,‘;r’:i::~-!~,~’~ :r ,,.
,I::,#!,,i, :,,~~,, ,,

SYNTHETIC
SEISMOGRAM

Fig. 18. Up-wave dephasedVSP, filtered to matchseismicdata.

Flg. 17. Primarycomposite VSP and synthetic seismogram.

in this area can be attributed to the presence of multiples,

which of course are absent on the VSP composite.
In order to correlate the VSP composites with the
surface seismic data, it is necessary to filter the broad-
band VSP to match the relatively narrower band seismic.
The result is shown in Figures 18 and 19, where it is
interesting to note that the deterioration in primary
energy at about I. I s on crossing the Wabamun level is
no longer as obvious as it was on the broad-band VSP
(Fig. 16). Apparently, the 23-Hz low-cut filter applied
to the seismic data discriminates against the relatively
low-frequency multiples, a result that is confirmed by
experienced seismic processors in the area. Fig. IS. Up-wavedephasedVSP, filtered to matchseismicdata
26 D.HAMPSONand
Fig. 20. Seismic data correlated with primary zone VSP composite.
Flg.21. SeismicdatacorrelafedwithVSPcomposite. recordedabove
the Wabamun.
Finally, Figures 20 and 21 show the correlation between
the two VSP composites and the surface sc:smic data.
The composite in Figure 20 is from the primary zone
and, in principle, contains no Wabamun-generated
multiples. The composite in Figure 21 contains the mul-
tiples and should match the seismic data better for this
reason. From about 0.8 s down, we see that this is the
case, and the event at about 0.85 scan be clearly identi-
tied as multiple because it appears on one composite
and not on the other. The most interesting feature is the
L.MEWHORT
event that appears at about 1.04 s on the left side of the
section and rises rapidly to intersect both composites at
around 1.01 s. The fact that it is much wcakcr on the
composite with multiples indicates that a multiple is
interfering with this event, and that effective multiple
attenuation may improve its continuity on the seismic
data.
MULTIPLE ATTENUATION
The interpretation of the VSP suggests that the major
multiple-generating mechanism is a simple one: the impor-
tant multiples are peg-legs generated between the
Wabamun and the near surface. As the multiple period
is constam, this should be an ideal case for a gapped
deconvolution. In practice, predictivcdccon has proved
ineffective in this area for two reasons: I) because the
Wabamun is so shallow, there arc usually only a small
number of inside traces where it is present without
interference from the first arrivals, and 2) because of
interference between the primaries and first-order
multiple, the design of the decon operator is unreliable.
The second problem can certainly be alleviated by using
the VSP. Figure 22 shows again the down-going wave
VSP. Since only multiples are present in this section, it
shouldbcpossible todcsignavcrycffectivcdcconvolution
operator. In order to use this operator on the seismic
data, an average operator was calculated by using the
time window from 0 to 800 ms and averaging over I6
traces as shown.
Note that these traces came from below the Wabamun
level. Thcoperatorlcngthis300 msand thclagis400 ms.
The effect of using this single operator on the down-
wave section is shown in Figure 23. It is especially
encouraging to note that the first-order multiple has
been signiticantly attenuated throughout the section
and not just within the design gate. The second-order
multiple between 1000 and 1100 ms has also been
attenuated, but a multiple between II00 and I200 ms is
still present. This is apparently a long period peg-leg
multiple set up between the deep primary reflectors and
the interval above the Wabamun. Figure 24 shows the
result of using the average operator on the up-wave
section. In this case, the operator has been applied only
to the traces recorded above the Wabamun. As expected,
the multiple has been strongly attenuated, while the
primary reflection below remains largely unaffected.
In using the derived operator on seismic data near the
VSP well, the assumption is that the change in wavc-
form that occurs as a result of the extra bounce through
the Wabamun is constant over an area and is indcpcn-
dent of the basic wavclct. The actual operator that is
used on each seismic trace can be thought ofas compris-
ing three components: the 300.ms operator from the
VSP, which contains information about the waveform
change; the lag time, which will vary according to posi-
tion on the seismic data, as well as offset; and the total
amplitude of the operator, which will depend on the
 VERTICAL SEfSMfC PROFILE DISTINGUISHES MULTIPLES 27
DESIGN GATE

DESIGN WINDOW

Fig. 22. Down-going waves before multiple anenuation

Fig. 23. Down-going waves afier predictive decon with average operator
28 D. HAMPSON
BEFORE DECON AFTER DECON
,, ,,
,; !Z ,!i ,I,,,
8, I), ’ Ji::~(( I,,’
‘, 1: ,‘I!, ;:,;;,1.,1~
uwuu .- .~-
IV/..,.,,~--..~
~w!pA,bmy.,,,~*lw,,tl;:!‘:’ #,,,1,,
i~.U4~~;::iil::r:.i*Y)Lj!/!jl;lil;,i
“:~!‘i!~r!~,:r~,~l!,)l~l’~li,,:,,.::,:’i,
w,.:.iJ~,:‘,~,,!~ ,,,)tI/,h,:[,t,I,,,)
,:,,,,,._._...,.,,i’ ~.Al.L ,,,.,
Fig.24. Useofaveragepredictivedeconoperatoronup-wavesection.
gain recovery used in processing the seismic data. The
approach used here was to find the optimum lag and
amplitude for each trace. The operator was applied to
common-depth-point datajust before stack and, as the
Wabamun reflection had been muted from many of the
far traces, a model was formed by stacking several
inside traces for input to the decon operation.
Figures 25 and 26 show the seismic data before and
after application of the VSP-derived operator. The pri-
mary zone composite has been inserted to help evaluate
the performance of the operator. Three important fea-
turesshouldbenoted. First, theevent marked”multiple”
is part of the first-order Wabamun multiple and has
been attenuated by the VSP operator. As a result. pri-
mary events between 700 ms and 900 ms arc now more
continuous. The second point is that the event marked
primary D is clearly more continuous after using the
attenuation operator. As indicated in the previous section,
it appears that a multiple had been generated from the
primary at about 550 ms on the seismic datafollowed by
the extra bounce between Wabamun and near surface.
This multiple was apparently obscuring primary D, and
the attenuation has improved the continuity of the
primary. The event ties primary D on the composite
trace closely enough to identify it, but the small mis-tie
between them implies some wavelet distortion caused
either by residual multiple interference or by the attenu-
ation method itself. The improvement in continuity is
especially obvious in Figures 27 and 28, which show a
portion of the seismic data at an enlarged scale. The
third point is that the multiples below the basement
have been significantly attenuated.
and L. MEWHORT
For the sake of comparison, a second multiple-
attenuation technique was tried on these data. The method
was similar to one proposed by Ryu (1980) in which F-K
filtering was used to eliminate multiples on the basis of
differential moveout. In the present case, the multiples
all had very similar moveouts at the time of interest, so
the procedure was:
I) NMO-correct the CDP-gathered data at the multi-
ple velocity.
2) Suppress events with close to zero-dip energy by
using F-K filtering.
3) NMO-correct the data at the primary velocity.
This procedure differs from Ryu’s in that only a small
segment of the F-K plane is atfenuated. rather than an
entire quadrant. As the multiple moveout is practically
constant over the zone of interest, the NM0 correc-
tions mentioned above amount to static time-shifts within
the zone of interest.
The resulting section is shown in Figure 29. In many
ways, the result is similar to the VSP attenuation. The
continuityofprimary Dhasagain beenimproved, although
it appears that the continuity of this event on the right
side of the anomaly is significantly better on the VSP-
attenuated section. The event marked “multiple” has
not been attenuated as well as on the VSP section.
Primaries A, B and C, on the other hand. appear quite a
bit stronger on the F-K filtered section.
In view of the better performance of the VSP-derived
operatorovcrthe F-Krnultipleattenuation,itisinterest-
ing to note the common offset stacks shown in Figure
30. Thecommonoffset stackisformed by adding together
sets of traces from nearby CDP locations with the same
shot-to-receiver offset. The result is a kind of average
CDP gather. The left-hand panel of Figure 30 shows the
stack before multiple attenuation, and the multiples can
be clearly identified by their residual normal moveout.
The common offset stack after the VSP multiple attenu-
ation shows that in fact the method has been most
effective on the near traces, while the residual multiples
are still strong on the far traces. Even though the F-K
method has apparently been more effective over the
whole range of offsets, the resulting stacked section
was shown to be inferior by using the VSP primary
composite to evaluate it. The conclusion is that the
near-tracemultiplesare mainlyresponsiblefortheinter-
ference on the stack section and that, by attenuating
them, the VSP-derived operator has produced an
improved stack.
On the whole, the two methods have performed com-
parably on this seismic line. Where the VSP has been
used to identify specific events. the VSP-derived opera-
torappears tohave performed slightly better. However,
it is not clear to what extent the derived operator can bc
used on seismic data shot at some distance from the
well. Inasmuch as the operator measures the change in
wavelet shape without specific reference to lag or input
wavelet shape, it is hoped that this quantity will be
reasonably constant over an extended area.
 “ERTlCAL SElSMlC PROFILE DISTINGUISHES MULTIPLES 29

4- PRIMARY A
C PRIMARY 6
+ PRIMARY C
C PRIMARY D

Fig. 25. Primary zone VSP compared with input seismic

 30 I). HAMPSON and I. MEWH”K,

PRIMARY
PRIMARY
PRIMARY
PRIMARY

Fig. 26. Primary zone VSP compared with seismic with VSP multiple removal.
 VERTICAL SElSMlC PROFILE DlSTlNGUlSHES MULTlPLES 31

Fig. 27. Primary zone VSP compared with input seismic.

Flg. 28. Primary zone VSP compared with seismic with VSP multiple removal.
 D. HAMPSON and L. MEWHORT

!8 Primq .zone VSP compared with seismic with F/K multiple removal.

NO MULTIPLE F/K MULTIPLE VSP MULTIPLE

ATTENUATION ATTENUATION ATTENUATION

Fig. 30. Common offset stacks

 VERTICAL SEISMIC PROFILE DISTINGUISHES MULTIPLES
CONCLUSIONS
The Vertical Seismic Profile is a tool that is well
suited to the study of the type of multiple problem that
exists in western Canada. In this study we have shown
that, bycompositingVSPtracestoproducetheprimties-
only and the primaries-plus-multiples responses, resid-
ual multiple effects could be unambiguously identified
on the seismic data. Information about the wavelet
shape from the down-wave VSP was used to design a
multiple-attenuationfilter, whichwasappliedwithsome
success to the surface-recorded data. A second multiple-
attenuation technique based on F-K filtering was tried
as well, and the VSP composite was used to evaluate
the result. Although the two methods achieved similar
results, it was felt that the VSP-derived operator per-
formed better on specific events.
Themajordiffkultyarisingfrom this VSPstudycame
from variations in the source wavelet. An improvement
33
in this area would have been achieved by using a more
repeatable source, such as “Vibroseis”* and monitor-
ing the swrce wavelet with a single reference geophone
at some depth from the shot.
* @Registeredtrademark and service mark of Conoco Inc.
REFERENCES
Hauge, P.S. 1981, Measurementsafattenuationfromve~iicalseismic
profiles: Geophysics,v. 46. p. 1548-1558.
Kennett, P., Ireson. R.L. and Coon. P.I. 1980, Venical seismic
profiles: their applications in exploration geophysics:Geophys.
Pros., v. 28, p. 676-699.
Ryu, J.V. 1980, Decomposition (DECOM) of seismic gathers into
velocity components by a space-time filter: Presented at the 50th
AnnualMeetingofSocietyofExplorationGeaphysicists(Houston).
Ziolkowski, A. and Lewill, W.E. 1979, A simple approach to high
resolution seismic profiling for coal: Geophys. Pros., v. 27, p.
m-393.